Bug 200748

Summary: Issues after resume when TSC timecounter selected, possibly due to CPU TSC resetting to zero
Product: Base System Reporter: Gavin Atkinson <gavin>
Component: kernAssignee: freebsd-bugs (Nobody) <bugs>
Status: New ---    
Severity: Affects Only Me    
Priority: ---    
Version: CURRENT   
Hardware: Any   
OS: Any   
Description Flags
Verbose dmesg none

Description Gavin Atkinson freebsd_committer freebsd_triage 2015-06-10 01:09:43 UTC
Created attachment 157592 [details]
Verbose dmesg

I have a Lenovo Flex 10 laptop.  With kern.timecounter.hardware=TSC (the default), on resume there are serious issues: primarily that time appears to run very fast, the AHCI controller/drive fail to reinit correctly, etc.  Unfortunately the video doesn't recover either (unrelated to this issue), so it's a bit hard to establish exactly what state the machine is in at this time.

After some digging, it looks like the CPU TSC counter is reset to zero on resume:

root@flex10:~ # cpucontrol -m 0x10 /dev/cpuctl0 ; cpucontrol -m 0x10 /dev/cpuctl1
MSR 0x10: 0x0000019a 0x3d2660a7
MSR 0x10: 0x0000019a 0x5f4b9d7a
root@flex10:~ # zzz
[wake machine up]
root@flex10:~ # cpucontrol -m 0x10 /dev/cpuctl0 ; cpucontrol -m 0x10 /dev/cpuctl1
MSR 0x10: 0x00000004 0x27c4514f
MSR 0x10: 0x00000004 0x49dcc993

I suspect (though haven't been able to prove) that this is the cause.

The CPU is an Intel N2807 ultra low power CPU.  It does have invariant P state TSC.